engineering, then management, now data science: 3 (4...
TRANSCRIPT
Engineering, then Management, now Data Science:
3 (4?) Stories of Intrapreneurship at the Faculty of Industrial Engineering & Management
Technion
Avishai Mandelbaum
IE&M & SEELab, Technionhttp://ie.technion.ac.il/Labs/ServEng/
• Caveat: I am an expert in neither Entrepreneurship nor Innovation
• But my faculty (school), Industrial Engineering & Management, carries within Technion the flag of “Management” (hence also the E&I Flag)
• And I happen to be the present Dean of IE&M
• Hence accepted the invitation/challenge
2
• Caveat: I am an expert in neither Entrepreneurship nor Innovation
• But my faculty (school), Industrial Engineering & Management, carries within Technion the flag of “Management” (hence also the E&I Flag)
• And I happen to be the present Dean of IE&M
• Hence accepted the invitation/challenge, encouraged by (my observation)o Management & Engineering – each is changing in the direction of the
other (research, teaching, practice) - if so, my experience possibly relevant: Managers forced to become more “analytic”: BI, big-data, … Engineers forced to “manage”: data-based research (A-journals),
teaching (e.g. case studies, reversed/active class-room), team work and personal E&I (within individual career-paths)
3
3 Stories:
Technion (50’s; State of Israel “born” in 1948)Industrial Engineering & Management (IE&M): born as Faculty & Profession
Personal (90’s; BSc=Math + CS, MSc = Stat, PhD = OR; 1st job = BusSchool)Service Engineering: Research, Teaching, Management Support
IE&M (Ongoing, both “IE” and “M” in flux)Data Science & Engineering (DS&E): BSc, Research (Culture)
4
4th Story?
Technion (50’s; State of Israel “born” in 1948)Industrial Engineering & Management (IE&M): born as Faculty & Profession
Personal (90’s; BSc=Math + CS, MSc = Stat, PhD = OR; 1st job = BusSchool)Service Engineering: Research, Teaching, Management Support
IE&M (Ongoing, both “IE” and “M” in flux)Data Science & Engineering (DS&E): BSc, Research (Culture)
… and perhaps we are now creating the 4th story?
5
1958 - “Born” at Technion, as a Department (and a Profession), by fusing • Manufacturing/Industrial Engineering, from Mechanical Engineering • Behavioral Sciences (Psychology, Sociology), from General Studies• Industrial Relations & Human-Resource Management, also from GS• Operations Research = newly established theme (included Statistics, Economics)
Founders: Prof’s Pinchas (Paul) Naor, Shmuel (Sam) Eilon, Sebastian Littauer.
6
Story 1: Industrial Engineering & Management (IE&M)
1958 - “Born” at Technion, as a Department (and a Profession), by fusing • Manufacturing/Industrial Engineering, from Mechanical Engineering • Behavioral Sciences (Psychology, Sociology), from General Studies• Industrial Relations & Human-Resource Management, also from GS• Operations Research = newly established theme (included Statistics, Economics)
Founders: Prof’s Pinchas (Paul) Naor, Shmuel (Sam) Eilon, Sebastian Littauer. Vision: All disciplines join and cross-nurture while still maintaining scholarly/academic legitimacy, via Areas: all educating IE&M’s while each awarding own MSc & PhD
1968 - Faculty (approx. School, or large department)• Later more areas: Economics, Statistics (70s), Information Systems (80s); • … and focus themes: Economics & Computation, Service Engineering, Entrepreneurship,
Innovation, …
7
IE&M: Brief History
“Product”: System/Industrial engineers, w/ sound scientific education, trained also in the functions of management (human resources, economics, marketing, finance, accounting)
8
IE&M: Brief History
“Product”: System/Industrial engineers, w/ sound scientific education, trained also in the functions of management (human resources, economics, marketing, finance, accounting)
Program in the 70’s: Industrial Management, turned later into MBA (now 3 programs), which is a main reason for not having a Business School at Technion
9
IE&M: Brief History
Most versatile and one of the largest of its kind world-wide• Profession constantly in high demand
o 11,000+ IE&M Technion alumnio 13 IE&M programs in Israel (research universities, colleges) o 1000 IE&M’s graduated in Israel last year (500, 500)
• World-class research institute o 50+ researchers, many world-class (editorial boards, prizes, …)o Guests, postdocs, ...o Alumni that are leading researchers in leading universities
IE&M: Founders’ Vision Materializes
• Versatility (Flexibility, Adaptability) = our strength
o 6 areas w/ critical mass in each: IE, OR, Stat, InfoSys, BehavSc+Management, Econo 13 Research Laboratories: Optimization, Business Intelligence, SEE,…
o Focus themes: Robust OptimizationBehavioral Economics,Economics & Computation, IS, Information Retrieval, NLP, Project Mgt/Edu, Stochastics, Game Theory, Service Engineering, …
IE&M: Founders’ Vision Materializes
Background: Math, CS, Stat, OR; Management, Engineering
History:
• 80’s (PhD) TQM Quality Math, Statistic • 90’s (Israel) BPR Processes OR, IE• 00’s (SEE) CRM Customers Service-Engineering, IS• 10’s BI/BD Data with all above
(Looking back: Did what I liked and fortunate that others liked as well)
12
Story 2: Service Engineering (& Science)
13
Research PartnersStudents:Aldor, Baron Yonit, Carmeli‐Yuviler Nitzan, Carmeli Boaz, Chen Hong, Cohen Izik, Feldman Zohar, Garnett, Gurvich, Khudiakov, Koren, Maman, Marmor, Reich, Rosenshmidt, Shaikhet, Senderovic, Tseytlin, Yom‐Tov, Zaied, Zeltyn, Zychlinski, Zohar Eti, Zviran, ...
Theory:Armony, Atar, Azriel, Cohen Izik, Gurvich, Feigin, Gal, Huang Junfei, Jelenkovic, Kaspi, Massey, Momcilovic, Reiman, Shimkin, Stolyar, Trofimov, Wasserkrug, Whitt, Yom‐Tov, Zeltyn, Zhang Jiheng, Zhang Hanqin, ...
Exploratory Data Analysis, Data Sources, Statistics, Projects:Brown, Gans, Shen Haipeng, Zhao Linda; Zeltyn; Ritov, Goldberg; Gurvich, Huang Junfei, Liberman;Liu Nan, Ye Han; Armony, Marmor, Tseytlin, Yom‐Tov; Gorfine, Ghebali; Tezcan; Kim Song‐Hee, Won Chul Cha; He Shuangchi, Sim Melvyn; Feigin, Azriel; Rafaeli, Yom‐Tov, …
Industry:Mizrahi Bank, Fleet Bank, Rambam Hospital, IBM Research, Hapoalim Bank,Pelephone Cellular, Samsung Hospital, Dana Farber Cancer Institute, LivePerson, …
Technion SEE Laboratory (SEELab):Feigin; Trofimov, Nadjharov, Gavako; Kutsy; Senderovic, Carmeli; Liberman, Koren, Plonsky; Research Assistants, Visitors, Postdocs, ...
15
16
Agents(CSRs)
Back‐Office
Experts
VIP)Training(
Arrivals(Business Frontier of the21th Century)
Redial(Retrial)
Busy(Rare)
GoodorBad
Positive: Repeat BusinessNegative: New Complaint
Lost Calls
Abandonment
Agents
ServiceCompletion
ForecastingStatistics,HumanResourceManagement(HRM)
New Services Design (R&D)Operations,Marketing,MIS
Organization Design:Parallel (Flat)Sequential (Hierarchical)Sociology, Psychology,Operations Research
HRM
Service Process Design
Quality
Efficiency Skill Based Routing (SBR) DesignMarketing, HRM,Operations Research,MISCustomers Interface
Design
Computer‐Telephony Integration ‐ CTIMIS/CS
Marketing
(Turnover up to200% per Year)(Sweat Shops
of the21th Century)
Operations/BusinessProcessArchiveDatabaseDesignData Mining:MIS, Statistics, Operations Research, Marketing
InternetChatmailFax
Lost Calls
Service Completion)75% in Banks(
( Waiting TimeReturn Time )
Logistics
Customers Segmentation ‐CRM
Psychology, Operations Research,Marketing
Expect 3 minWilling 8 minPerceive 15 min
PsychologicalProcessArchive
Psychology,Statistics
TrainingJob Enrichment
Marketing, Psychology,Operations Research
Human Factors Engineering
VRU/IVR
Queue)Invisible(
VIP Queue
(If Required 15 min,then Waited 8 min)(If Required 6 min, then Waited 6 min)
Information Design
FunctionScientific DisciplineMulti‐Disciplinary
Index
Tele‐StressPsychology
Operations Research,Economics, HRM
Game Theory,Economics
Incentives
Service Engineering of a Telephone Call Center(Esprit de Corps)
Redial
ImagingLaboratory
Experts
Interns
Returns (Old or New Problem)
“Lost” (LWBS) Patients
LWBS
Nurses
Statistics,HumanResourceManagement(HRM)
New Services Design (R&D)Operations,Marketing,MIS
Organization Design:Parallel (Flat) = ERvs. a true ED Sociology, Psychology,Operations Research
Service Process Design
Quality
Efficiency
CustomersInterface Design
Medicine
(High turnoversMedical‐Staff shortage)
Operations/BusinessProcessArchiveDatabaseDesignData Mining:MIS, Statistics, Operations Research, Marketing
StretcherWalking
Service Completion(sent to other department)
( Waiting TimeActive Dashboard )
PatientsSegmentation
Medicine,Psychology, Marketing
PsychologicalProcessArchive
Human FactorsEngineering(HFE)
InternalQueue
OrthopedicQueue
Arrivals
FunctionScientific DisciplineMulti‐Disciplinary
Index
ED‐StressPsychology
OperationsResearch, Medicine
Service Engineering of an Emergency Department
Returns
TriageReception
Skill Based Routing (SBR) DesignOperations Research,HRM, MIS, Medicine
IncentivesGame Theory,Economics
Job EnrichmentTrainingHRM
HospitalPhysiciansSurgical
Queue
Acute,Walking
Blocked(Ambulance Diversion)
Forecasting
Information DesignMIS, HFE,Operations Research
Psychology,Statistics
Home
20
2007-Donation: Hal and Inge MarcusFounders: Feigin P., Mandelbaum A.Researches: Valery Trofimov, Ella Nadjharov, Igor Gavako, …
• Students• Guests
Technion SEE LabService Enterprise Engineering / Science
SEELab: Environment for graphical EDA in real-time
21
Detailed operational histories (customers, servers), e.g.
1. *Bank Anonymous : 1 year, 350K calls by 15 agents - in 2000,which paved the way to:
2. *U.S. Bank : 2.5 years, 220M calls, 40M by 1000 agents3. Israeli Cellular: 2.5 years, 110M calls, 25M calls by 750 agents4. Israeli Bank: from January 2010, daily-deposit at a SEESafe5. Service Engineering internet site: click-stream data (2 years)
6. *Home (Rambam) Hospital : 4 years, 1000 beds, inter-ward flow
7. Emergency Department (ED) patient flow:• 5 EDs in Israel: 1-2 years, late David Sinreich, ED arrivals & LOS• ED in Seoul: 2 months, K. Song-Hee & W. Cha, pilot• ED in XY: 2 years, pilot
8. Hospital RTLS (Real-Time Location System):• 250K events/day: 1000 patients, 350 staff (1500 tagged entities)• Infrastructure: 900 readers (sensors), many floors
*Open & Free for research and teaching
Agents(CSRs)
Back‐Office
Experts
VIP)Training(
Arrivals(Business Frontier of the21th Century)
Redial(Retrial)
Busy(Rare)
GoodorBad
Positive: Repeat BusinessNegative: New Complaint
Lost Calls
Abandonment
Agents
ServiceCompletion
ForecastingStatistics,HumanResourceManagement(HRM)
New Services Design (R&D)Operations,Marketing,MIS
Organization Design:Parallel (Flat)Sequential (Hierarchical)Sociology, Psychology,Operations Research
HRM
Service Process Design
Quality
Efficiency Skill Based Routing (SBR) DesignMarketing, HRM,Operations Research,MISCustomers Interface
Design
Computer‐Telephony Integration ‐ CTIMIS/CS
Marketing
(Turnover up to200% per Year)(Sweat Shops
of the21th Century)
Operations/BusinessProcessArchiveDatabaseDesignData Mining:MIS, Statistics, Operations Research, Marketing
InternetChatmailFax
Lost Calls
Service Completion)75% in Banks(
( Waiting TimeReturn Time )
Logistics
Customers Segmentation ‐CRM
Psychology, Operations Research,Marketing
Expect 3 minWilling 8 minPerceive 15 min
PsychologicalProcessArchive
Psychology,Statistics
TrainingJob Enrichment
Marketing, Psychology,Operations Research
Human Factors Engineering
VRU/IVR
Queue)Invisible(
VIP Queue
(If Required 15 min,then Waited 8 min)(If Required 6 min, then Waited 6 min)
Information Design
FunctionScientific DisciplineMulti‐Disciplinary
Index
Tele‐StressPsychology
Operations Research,Economics, HRM
Game Theory,Economics
Incentives
Service Engineering of a Telephone Call Center
Redial
23
24
25
26
Use Case: Skills-Based Routing in a Call Center
• Customer Classes– Marketing segregates customers according to their needs and/or importance – this
determines customer priorities
• Agent Skills– Human-Resources Management assigns agent skills according to capabilities,
experience (training) – this determines agent constituencies
• Matching Class & Skill (Demand and Supply)– Operations-Researchers develop matching algorithms so that customers don’t wait long
for an agent (service-level) and agents don’t wait long for a customer (efficiency)
• Information Infrastructure (IS/CS)
• Data Management (Statisticians)
28
29
Building Blocks of a Model: Service Durations
30
Durations: Phone Calls (Surprises)
31
ImagingLaboratory
Experts
Interns
Returns (Old or New Problem)
“Lost” Patients
LWBS
Nurses
Statistics,HumanResourceManagement(HRM)
New Services Design (R&D)Operations,Marketing,MIS
Organization Design:Parallel (Flat) = ERvs. a true ED Sociology, Psychology,Operations Research
Service Process Design
Quality
Efficiency
CustomersInterface Design
Medicine
(High turnoversMedical‐Staff shortage)
Operations/BusinessProcessArchiveDatabaseDesignData Mining:MIS, Statistics, Operations Research, Marketing
StretcherWalking
Service Completion(sent to other department)
( Waiting TimeActive Dashboard )
PatientsSegmentation
Medicine,Psychology, Marketing
PsychologicalProcessArchive
Human FactorsEngineering(HFE)
InternalQueue
OrthopedicQueue
Arrivals
FunctionScientific DisciplineMulti‐Disciplinary
Index
ED‐StressPsychology
OperationsResearch, Medicine
Service Engineering of an Emergency Department
Returns
TriageReception
Skill Based Routing (SBR) DesignOperations Research,HRM, MIS, Medicine
IncentivesGame Theory,Economics
Job EnrichmentTrainingHRM
HospitalPhysiciansSurgical
Queue
Acute,Walking
Blocked(Ambulance Diversion)
Forecasting
Information DesignMIS, HFE,Operations Research
Psychology,Statistics
Home
33
34
35
36
Erlang-R Fluid Model
37
38
FNet vs. Data :Erlang-R value
39
40
41
Story 3: What’s Next - Diagnosis
• Many fast changes, mainly due to technological advances but no less so because of data availability through these technologies: unprecedented quantity, granularity and quality
• Strong market demand for “data-professionals” (vaguely defined)
• IE&M has always had a “Data Culture” (Climate): in research and teaching, collected in the real-world or research labs (controlled experiments)
• Technion & IE&M, being research-driven institutions, advance the boundary of knowledge, and disseminate the knowledge created through teaching and alumni
• IE&M has been a great success-story: 13 programs in Israel is clear testimony to our success, but it also calls for perhaps Reinvention or at least Invigoration of the profession (within Technion) – and we are in the process of doing both
What’s Next: Prognosis
• Reinvention: Created the profession Data Science & Engineering - B.Sc. in DS&E, who will extract useful information from vast amounts of data via computerized technologies
• Invigoration: DS&E requires additional faculty members, new courses and improved infrastructure, which will be used to invigorate the existing IE&M (e.g. Increase fraction of electives, consolidate concentrations)
Data Handling (Education)
Gathering Managing Analyzing Visualizing
• Mechanism design (crowd sourcing)
• Event processing and sensors (hospitals, transportation)
Tasks
• Data integration• Big graph partitioning• Probabilistic data
management
• (Sponsored) Search• Sentiment analysis• Decision making• Time series• Process analysis• Ad exchange• Differential pricing• Prediction
Areas/Tools
• Web and social media crawling
• Stream processing• Information elicitation
• Search engines• Databases (SQL,
noSQL, newSQL, probabilistic DB)
• Cloud computing
• Economics & computation• Information retrieval & natural language processing• Operations research (optimization, simulation)• Statistics & machine learning
• User experience• Cognitive science
• Interaction• Animation
46
Gaining Experience with Data:
Data collection and management laboratory Data analysis and presentation laboratory Industry project (summer internship)
Data Types:
Data collection and management:
Database management systems Managing distributed data Distributed information systems
Information analysis and presentation:
Statistical theory Data mining Computational learning and online optimization Visual information presentation and cognition
Scientific and Engineering Fundamentals:
Computer Science:
Introduction to computer science
Data structures and algorithms Introduction to computation
and logic modeling
General:
Physics 1 Two scientific courses Physical education English
Mathematics:
Two calculus courses Algebra Discrete mathematics Numerical analysis Probability Rational agents
Data Science and Engineering
Introduction to data science Deterministic models in operations
research Stochastic models in operations
research
Introduction to statistics Fundamentals of artificial
intelligence Models of electronic
commerce
Industrial psychology Human factor
engineering Numerical simulation
Applied econometrics (economic) Information retrieval (textual) Natural language processing
(textual)
Service system engineering (operational)
Periodic series (time-based) Event processing (sensory)
Environmental data Epidemiological data
47
Closing the Data-Gap: Call-Centers & Hospitals
• Large call center:
o Thousands of agents
o Hundreds of thousands of calls per day
o Automatically, individual customers & agents (event log)
• Operational
• Financial, ...
• Large hospital:
o 1000+ Beds
o Thousands of nurses, hundreds of doctors
o Incomplete data (at best) ...
50
51
52
Statistical‐Deviation of Planned vs. Actual
58
59
Applications in DFCI
60
Control: rooms status, physicians location, long wait times
Planning: number infusion chairs, load-balancing among floors
Management: evidence-based
Design vs. Performance: exam durations (no 15-minutes)
Motivating improvement: room for physician vs. room for patient
e.g. Specific Emergency-Department, with ample reliable data
• Real-time: data-based control of patient-flow (bottlenecks)
• Short-term: on Monday, set Tuesday’s staffing levels (or next week’s)
• Long-term: calculate real cost of care for individual patient (as opposed to insurance costs)
• Above, by online creation of data-based ED model(s) (empirical, simulation, mathematical)
61
SEELab: Converging to the Vision
62
63